Method and system for detecting liquid level height of liquid silicon by using target spot reflection

Abstract

The invention discloses a liquid silicon liquid level height detection method using target spot inversion, which comprises the following steps: arranging a graphite ring above the liquid silicon liquid level, and installing a target spot on the graphite ring; obtaining a target spot and an inverted image formed by the target spot on the liquid level of the liquid silicon; the acquired image is preprocessed; performing first deep learning processing on the preprocessed image to obtain a coarse region calibration image of the target spot and the inverted image of the target spot; performing second deep learning processing on the calibrated coarse region image to obtain a pixel distance between the target spot and the inverted image of the target spot; and determining the actual distance between the target spot and the liquid level of the liquid silicon according to the pixel distance. The influence of liquid level fluctuation and crystal rotation on the target spot is small, the detection precision and reliability are improved, the anti-interference capability is remarkably enhanced, and a key technical guarantee is provided for the monocrystalline silicon growth process.

Description

Technical Field

[0001] The invention relates to the field of single crystal silicon growth, and in particular to a liquid silicon level detection method and system using target point reflection. Background Art

[0002] During the growth of single crystal silicon, as the crystal grows, liquid silicon gradually adheres to the silicon rod to form solid silicon. As the liquid silicon in the crucible decreases, the liquid level will gradually decrease. Ensuring a stable liquid surface distance is crucial for crystal growth. If the liquid level is too low, thinning and wire breakage will occur. If the liquid level is too high, there is a risk of furnace explosion. In order to ensure that the relative position between the liquid level and the bottom of the silicon rod is fixed, a servo motor is required to lift the crucible. The moving distance of the crucible is determined by a visual inspection system.

[0003] The mainstream liquid surface distance detection method in the market is to use traditional machine vision to detect the width of the reflection of the graphite ring on the liquid surface, and then estimate the liquid surface distance based on the reflection width. However, this method is easily affected by liquid surface fluctuations and crystal rotation, resulting in low accuracy and poor stability. Summary of the invention

[0004] Purpose of the invention: In view of the above shortcomings, the present invention provides a high-precision, strong anti-interference method and system for detecting the liquid silicon liquid level using target reflection.

[0005] Technical solution: To solve the above problems, the present invention adopts a liquid silicon level detection method using target reflection, including setting a graphite ring above the liquid silicon level, including:

[0006] Mounting the target on the graphite ring;

[0007] Acquire an image of the target and its reflection on the liquid silicon surface;

[0008] Preprocessing the acquired images;

[0009] The preprocessed image is processed by a first deep learning process to obtain a rough area calibration image of the target and its reflection;

[0010] After the calibrated coarse area image is processed by the second deep learning, the pixel distance between the target point and its reflection is obtained;

[0011] The actual distance between the target point and the liquid silicon surface is determined according to the pixel distance.

[0012] Furthermore, the preprocessing of the acquired image includes cropping the acquired image, adjusting brightness, enhancing image edges, and normalizing image data.

[0013] Furthermore, the coarse area calibration image of the target and its reflection is cropped at the calibration position to obtain a cropped coarse area image, the cropped coarse area image is normalized, and the normalized coarse area image is processed by a second deep learning process.

[0014] Furthermore, the color of the target point has a high contrast with the color of the graphite ring.

[0015] Furthermore, the image data in the single crystal furnace is acquired in real time by a high-resolution image acquisition device, and the image data includes the target point and the reflection of the target point on the liquid silicon surface.

[0016] The present invention also adopts a liquid silicon liquid level detection system using target point reflection, comprising:

[0017] An image acquisition module is used to acquire an image of a target point on the graphite ring above the liquid silicon surface and a reflection of the target point on the liquid silicon surface;

[0018] A preprocessing module, used for preprocessing the acquired image;

[0019] A coarse area determination module is used to obtain a coarse area calibration image of a target and its reflection after processing the preprocessed image through a first deep learning process;

[0020] A pixel distance determination module is used to obtain the pixel distance between the target point and the reflection of the target point after the calibrated coarse area image is processed by the second deep learning;

[0021] The actual distance determination module is used to determine the actual distance between the target point and the liquid silicon surface according to the pixel distance.

[0022] Beneficial effect: Compared with the prior art, the significant advantage of the present invention is that the target is less affected by liquid level fluctuations and crystal rotation, which improves the accuracy and reliability of detection and significantly enhances the anti-interference ability, providing key technical guarantees for the single crystal silicon growth process. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the process of the height detection method in the present invention. DETAILED DESCRIPTION

[0024] like Figure 1As shown, a method for detecting the liquid silicon liquid level using a target reflection in this embodiment is used. This method installs a target on a graphite ring and detects the reflection of the target on the liquid surface through a visual system. Combining the spatial position of the target and the position of the reflection, the precise liquid surface distance can be calculated. This method has the characteristics of high precision and strong anti-interference, has very high practical value, and can provide stable and reliable liquid level data for crystal growth. It includes the following steps:

[0025] Step 1: Install the target on the graphite ring;

[0026] Step 2: The image data in the single crystal furnace is acquired in real time by a high-resolution image acquisition device, and the image data includes the target point and the image of the target point reflected on the liquid silicon surface;

[0027] Step 3: Preprocess the acquired image, including accurately cutting the target moving area and capturing a specific range of images containing the target moving trajectory. The captured image is further preprocessed, mainly including brightness adaptive adjustment, image edge feature enhancement, and image data normalization.

[0028] Step 4: Input the pre-processed image data into the first deep learning model that has been pre-trained. The model uses advanced deep learning algorithms to quickly and accurately identify the rough position of the target and obtain a rough area calibration image of the target and its reflection.

[0029] Step 5: The coarse area calibration image of the target and the target reflection is cropped at the calibration position to obtain the cropped coarse area image, and the cropped coarse area image is normalized to further optimize the data quality and improve the accuracy of subsequent analysis. The image data after secondary normalization is input into another pre-trained second deep learning model. The model, through its high-precision recognition ability, finally outputs the precise location information of the target and obtains the pixel distance between the target and the target reflection.

[0030] Step 6: Determine the actual distance between the target and the liquid silicon surface based on the pixel distance to achieve liquid level detection.

Claims

1. A method for detecting the liquid silicon level using target reflection, comprising arranging a graphite ring above the liquid silicon level, characterized in that: include: Mounting the target on the graphite ring; Acquire an image of the target and its reflection on the liquid silicon surface; Preprocessing the acquired images; The preprocessed image is processed by a first deep learning process to obtain a rough area calibration image of the target and its reflection; After the calibrated coarse area image is processed by the second deep learning method, the pixel distance between the target point and its reflection is obtained; The actual distance between the target point and the liquid silicon surface is determined according to the pixel distance.

2. The method for detecting the liquid silicon level according to claim 1, characterized in that: The preprocessing of the acquired image includes cropping, brightness adjustment, image edge enhancement and image data normalization of the acquired image.

3. The method for detecting the liquid silicon level according to claim 2, characterized in that: The coarse area calibration image of the target and its reflection is cropped at the calibration position to obtain a cropped coarse area image, the cropped coarse area image is normalized, and the normalized coarse area image is processed by a second deep learning process.

4. The method for detecting the liquid silicon level according to claim 1, characterized in that: The color of the target point has a high contrast with the color of the graphite ring.

5. The method for detecting liquid silicon level according to claim 1, characterized in that: The image data in the single crystal furnace is acquired in real time through high-resolution image acquisition equipment, and the image data includes the target point and the reflection of the target point on the liquid silicon surface.

6. A liquid silicon level detection system using target reflection, characterized in that: include: An image acquisition module is used to acquire an image of a target point on the graphite ring above the liquid silicon surface and a reflection of the target point on the liquid silicon surface; A preprocessing module, used for preprocessing the acquired image; A coarse area determination module is used to obtain a coarse area calibration image of a target and its reflection after subjecting the preprocessed image to a first deep learning process; A pixel distance determination module, used to obtain the pixel distance between the target point and the reflection of the target point after the calibrated coarse area image is processed by the second deep learning; The actual distance determination module is used to determine the actual distance between the target point and the liquid silicon surface according to the pixel distance.

7. The liquid silicon level detection system according to claim 6, characterized in that: The preprocessing of the acquired image includes cropping, brightness adjustment, image edge enhancement and image data normalization of the acquired image.

8. The liquid silicon level detection system according to claim 6, characterized in that: The actual distance determination module is used to perform calibration position cropping on the coarse area calibration image of the target and its reflection to obtain a cropped coarse area image, normalize the cropped coarse area image, and perform a second deep learning process on the normalized coarse area image.

9. The liquid silicon level detection system according to claim 6, characterized in that: The color of the target point has a high contrast with the color of the graphite ring.

10. The liquid silicon level detection system according to claim 6, characterized in that: The image acquisition module acquires image data in the single crystal furnace in real time through a high-resolution image acquisition device, and the image data includes the target point and the reflection of the target point on the liquid silicon surface.

Images